Use cases

Six workflows a working PA can run today.

The connector earns its place when it removes a real piece of work. These are six concrete prompts a public adjuster can paste into Claude Desktop or ChatGPT today, with the underlying tool calls the assistant will make and the kind of answer it produces.

Real workflows.

Tool calls below reference the read tools in the tool reference. The assistant decides which tools to call based on the question; the sequences below are typical, not prescriptive.

Summarize all open claims with deadlines this week.

Monday morning. The firm has 40 active claims. The operator wants the week-ahead view without opening 40 files.

Typical tool sequence

  • search_claims — Resolves the operator's open workload.
  • get_claim_tasks — Pulls open tasks with due dates per claim.
  • get_claim_readiness — Surfaces blockers that look like deadlines (statute windows, response cutoffs).

Outcome

The assistant returns a grouped list: claim, deadline, what's due, who owns it. The operator picks the three most urgent and starts the day on those.

Draft a status update for Mrs. Reyes based on the timeline since last Wednesday.

The client has called twice asking for an update. The adjuster wants something accurate, dated, and short before responding.

Typical tool sequence

  • search_claims — Resolves 'Reyes' to the right claim.
  • get_claim_timeline — Reads everything that's happened in the last seven days.
  • get_claim_communications — Confirms the most recent inbound and outbound.

Outcome

The assistant writes a short update grounded in real events with dates, naming the carrier action and the next step. The operator reviews, edits, and sends it from their own email.

Which claims are waiting on a carrier reply for more than seven days?

Mid-week triage. The operator wants to escalate or follow up on stalled negotiations before they age past statute windows.

Typical tool sequence

  • search_claims — Returns active claims.
  • get_claim_communications — Reads the latest direction and gap on each claim.
  • get_claim_participants — Surfaces the desk adjuster's contact info for the ones that need follow-up.

Outcome

The assistant returns a sorted list of claims by days-since-last-carrier-response, with the desk adjuster's phone and email pre-filled. Two minutes of triage instead of an hour.

What changed across all my claims since yesterday?

End-of-day check-in or first thing in the morning. The operator wants to know what moved without scrubbing the timeline of every file.

Typical tool sequence

  • search_claims — Returns the operator's active claims.
  • get_claim_timeline — Pulls the most recent entries on each, filtered to the last 24 hours.

Outcome

The assistant returns a per-claim digest: new inbound communications, status changes, uploaded files, completed tasks. The operator sees the day's movement on one screen.

Pull a chronological summary of the Patel claim with everything that would matter at deposition.

An attorney wants the file ahead of a deposition. The PA wants to send a clean summary with cited evidence rather than 200 pages of raw timeline.

Typical tool sequence

  • get_claim_snapshot — Anchors facts: parties, dates, carrier, status.
  • get_claim_timeline — Reads the full chronology.
  • get_policy_excerpts — Pulls operator-approved policy language for any coverage dispute.
  • search_claim_evidence — Finds the carrier's denial letter and any supporting estimates.
  • get_evidence_chunk — Returns the source text for direct citation.

Outcome

The assistant returns a dated narrative with direct quotes from approved policy excerpts and citation ids back to the underlying files. The PA reviews, attaches the source files, and sends the summary.

What evidence is missing on the Wong roof claim before we can send the carrier package?

Pre-package check. The operator wants the readiness gap before assembling the package, not while assembling it.

Typical tool sequence

  • get_carrier_package_readiness — Reads the package-readiness advisory.
  • get_file_workspace_summary — Surfaces files in review or extraction-incomplete states.
  • get_claim_weather_report — Confirms weather evidence is attached for the date of loss.

Outcome

The assistant returns a precise checklist: 'Roof elevation photos missing for east side. Mitigation invoice not yet OCR'd. Weather report attached, hail size 1.75 in., NOAA confirmed.' The operator works the checklist instead of guessing.

A note on what the assistant won't do.

The connector returns data; the assistant composes the answer. The assistant will not send emails, mark anything resolved, change a claim status, or move money. With claim:write it can log a task or append a timeline note, both of which the operator can edit or close in the normal UI.

A draft is a draft. When the assistant produces a status update or a demand-letter outline, that text lives in the assistant's chat until the operator copies it into claimOS, an email, or wherever it needs to go. The operator stays accountable for what gets sent.